Alternative strategies for supply air temperature control in office buildings

Alternative strategies for supply air temperature control in office buildings

Energy and Buildings 82 (2014) 406–415 Contents lists available at ScienceDirect Energy and Buildings journal homepage: www.elsevier.com/locate/enbu...

2MB Sizes 1 Downloads 99 Views

Energy and Buildings 82 (2014) 406–415

Contents lists available at ScienceDirect

Energy and Buildings journal homepage: www.elsevier.com/locate/enbuild

Alternative strategies for supply air temperature control in office buildings Mattias Gruber ∗ , Anders Trüschel, Jan-Olof Dalenbäck Chalmers University of Technology, Energy and Environment, Building Services Engineering, Sweden

a r t i c l e

i n f o

Article history: Received 2 April 2014 Received in revised form 4 June 2014 Accepted 25 June 2014 Available online 28 July 2014 Keywords: Heating, ventilation and air-conditioning Building automation Supply air temperature control Optimal control Office buildings Genetic algorithm Energy efficiency Indoor air quality Thermal comfort

a b s t r a c t A key element for a reduced energy usage in the building sector is to improve the systems for indoor climate control. But, it is important that such measures are fairly simple and easy to implement in order to facilitate a widespread utilization. In this work, four alternative strategies for supply air temperature control in offices were investigated through simulations. Their level of complexity stretches from linear SISO (single-input, single-output) structures with standard inputs to an optimal algorithm with information about the entire set of disturbances acting on the building. The study was conducted with a conventional outdoor-air-temperature based method as benchmark, and two different heating, ventilation and air-conditioning (HVAC) systems as well as two types of building structures were taken into account. Compared to the benchmark, all alternative strategies resulted in lower energy usages while thermal comfort and indoor air quality were satisfied, and simple strategies could perform almost equally well as more complex. But, it was also shown that the benefits were highly dependent on the considered HVAC system and somewhat dependent on the considered building structures. © 2014 Elsevier B.V. All rights reserved.

1. Introduction Approximately 40% of the global final energy is used in buildings, and in the European countries, 76% of this share goes towards systems for heating, ventilation and air-conditioning (HVAC). The purpose of an HVAC system is to achieve a desirable indoor climate and a key element for reducing the associated energy usage is to improve the building automation system. Many systems for indoor climate control consist of both local (on room level) and central components, and since their functions in many cases coincide, an important aspect in this context is to synchronize their operation. [1] An established technology for ventilation system control is DCV (demand controlled ventilation). The supply air flow rate to each space is then controlled individually and a common application is to use zone measurements (typically temperature or CO2 ) as input to a feed-back loop. The supply air temperature (SAT) is in turn first and foremost controlled centrally which means that the level should suite the entire range of demands within the building zones simultaneously. Conventionally, SAT control strategies are often based

∗ Corresponding author. Tel.: +46 317721168; fax: +46 317721152. E-mail addresses: [email protected], [email protected] (M. Gruber). http://dx.doi.org/10.1016/j.enbuild.2014.06.056 0378-7788/© 2014 Elsevier B.V. All rights reserved.

on external disturbances, and typically the outdoor air temperature (OAT) since the whole building is affected by its variations. This means that the SAT is increased for decreased OATs and vice versa [2]. But, such strategy can be problematic in office buildings since internal disturbances (such as people, lighting and equipment) are commonly dominating over the external climate during working hours. Then, a cooling demand can be expected almost regardless of the OAT (especially in modern office buildings with tight and well insulated envelopes), at the same time as the quality of the indoor climate is of utmost importance. Unsuitable supply air temperatures in certain zones can partly be compensated for by local components, but with an increased energy usage as a consequence. However, in more severe cases, the function of the HVAC system can be compromised. An emerging technology for achieving a SAT control that coincides with the remaining functions in DCV systems is to treat the HVAC operation as a global optimization problem by involving various objectives and variables. For example, an entire building’s energy usage can be minimized at the same time as the quality of the indoor climate is ensured. This approach has been studied in several previous works, as for example in [3], where the solver genetic algorithm (GA) was used in an office environment during two summer weeks by incorporating thermal climate and energy for air-handling in the objective function. A similar investigation was presented in [4], and it was found that up to 30% of

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

Nomenclature C c cp E ˙ c˙ M, Q Q˙ t V V˙ W ˙ W

thermal capacity [J/◦ C] CO2 concentration [ppm] specific heat capacity [J/(kg K)] total energy, weighted sum of energy types [W] CO2 flow rate [ml/s] thermal energy [J] thermal power [W] Celsius temperature [◦ C] volume [m3 ] volume flow rate [m3 /s] electric work [J] electric power [W]

Greek letters  time [s]  density [kg/m3 ] Subscripts max maximum min minimum r room s supply tot total vent ventilation

the energy could be saved by employing GA for SAT control in an educational environment instead of maintaining a constant level. These findings are moreover consistent with the reported results of [5] where genetic fuzzy optimization was used for climate control in an academic building scale model. Compared to maintaining constant SAT levels between 12 and 14 ◦ C, the strategy yielded energy savings between 54 and 61% during summer and winter ambient conditions, and this range was furthermore assessed as equivalent to savings between 29 and 36% on an annual basis. A final example of an investigation considering SAT control through global optimization algorithms is [6] in which GA was evaluated for ventilation system control in a multi-zone simulation platform. Simplified adaptive control models of the process were used to estimate the response to various external and internal conditions while the solver searched for a trade-off between air-handling energy and comfort aspects. Simulations were conducted for four different weather conditions and it was found that up to about 40% of the energy could be saved using the optimal approach compared to maintaining constant setpoints. A bit different approach to the same problem was in turn chosen in [7]. Instead of using numerical tools, expressions for the HVAC energy usage were formulated for a number of conditions and the optimal SAT was derived analytically. The theory was applied on a ventilation system in an office building and an energy saving potential between 8 and 27% compared to constant setpoint cases was indicated. 1.1. Background and purpose In the light of the previously cited works, it is clear that optimal SAT control strategies have a large potential for reducing the HVAC energy usage while maintaining a desirable indoor climate. However, a major drawback is that the associated level of complexity normally is high since a complete model of the process (building and HVAC system) is required together with extensive information about internal and external conditions. The process model is used to predict the influence of choosing a certain SAT under the present conditions, and an optimization algorithm

407

typically performs a search to find the most appropriate level. Even though physical models have the potential of describing the process sufficiently accurate, a large number of parameters would then be required whereof several are uncertain or hard to determine. Another alternative is black-box models that are constructed from observed data in terms of input/output measurements. But since the accuracy immediately becomes uncertain when operated in a range from where the training data lack information, the commissioning phase can be an extensive process. When it comes to gathering information about internal and external conditions, the customary set can be divided into a number of disturbances whereof several could be determined using established measurement technology (for example OAT, lighting, solar radiation). But for others, this is not an option (considering people and infiltration flow rate for example) without involving models for predictions and/or estimations, and these are typically afflicted with the same problem as the process models [8]. The aim of the present paper is to contribute in the search for simple SAT control strategies with a potential of reducing energy usage while achieving a desirable indoor climate. The investigation was performed in a simulated office environment and the core consists of three SISO (Single-Input, Single-Output) strategies that primarily are based on information inside the building through standard inputs. While a conventional OAT-based strategy was included as a benchmark, the upper boundary of potential energy savings was represented by an optimal algorithm, and it was furthermore examined if patterns in the solution could be utilized to formulate a more general and simple strategy. During simulations of a 11 room office building over two working weeks of Swedish summer and winter respectively, all strategies were individually evaluated for generating central SAT setpoints in two different DCV-HVAC system applications: an allair system with re-heaters in each room, and a system with hygienic ventilation and local water-based cooling and heating. In both cases, the thermal climate and indoor air quality (IAQ) within the zones were controlled with temperature and CO2 as indicators, respectively. The investigation was moreover conducted for different variants of the building and the results are presented as the total energy used by the HVAC system during the simulated periods. This paper is organized as follows: first, the simulation platform, including HVAC system and building, is presented together with its considered variants and simulated conditions. This is followed by a description of the evaluated SAT control strategies, and finally, the results from the study are presented and discussed. 2. Simulation framework The investigation is based on simulations performed in MATLAB® Simulink® and the platform consists of a building and an HVAC part. These are in turn made up from a large number of component models with one-dimensional equations used to describe how the simulated variables vary over the platform. The controlled variables of the rooms (CO2 and temperature) were calculated by physical balance equations and the relations between pressure and flow in the HVAC system were based on empirical data. In this section, the modelling approach is first described briefly and if more detailed information is desired reference [9] or [10] can be considered. Further on, the variants of the building and HVAC system are introduced which is followed by a description of the disturbances that were considered. 2.1. Building platform The building part is presented in Fig. 1. It was modelled to represent part of a floor in a modern office building with tight and well

408

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

Fig. 1. Layout of the modelled building with associated duct system for supply and exhaust air ventilation.

insulated outside walls, and consists of 11 individual rooms that are thermally connected to its neighbours. Most (9) rooms were modelled as cellular offices with a maximum occupancy of one person and a floor area of 10 m2 . While three of the consecutive walls are facing adjacent rooms, one is external and 80% of its surface is covered by a window with outside solar shadings. The platform also contains a corridor and a meeting room, both without external walls, and with floor areas of 140 and 18 m2 respectively. All rooms except the corridor have individual supply and extraction of ventilation air via the duct system depicted by arrows in Fig. 1. Air can also be directly exchanged between the rooms via open doors and the corridor, while leakage through internal walls is neglected. All rooms have adiabatic boundary conditions on roof and floor and the temperature of building elements (BE) such as walls, floors and roofs were calculated in two nodes located on the inside and outside surfaces. The heat exchange between a certain building element (i) and the surroundings is described by two coupled ODEs (ordinary differential equation) as presented in Eq. (1a). The room air temperatures (tr ) were calculated in one node per room using the ODE in Eq. (1b). The right hand side of this equation consists of terms describing the heat exchange between the air in a certain room and the building elements (Q˙ BE,i→r − Q˙ r→BE,i ) and terms describing the heat emitted by internal disturbances (Q˙ D,i→r ). Also the air CO2 concentrations (cr ) were calculated in one node per room using the mass-balance in Eq. (1c). The terms on the right hand side describe the CO2 exchange between a certain room and the ambience (c˙ →r − c˙ r→ ) as well as the CO2 emitted by internal ˙ D,i→r ). The building model and its variants were disturbances (M tested and validated against IDA ICE which in turn has been tested in the Building Energy Simulation Test (BESTEST) developed under IEA SHC Tasks 8, 12 and 22. dtBE,i × CBE,i = Q˙ →BE,i − Q˙ BE,i→ d

(1a)

dtr × Cr = Q˙ BE,i→r − Q˙ r→BE,i + Q˙ D,i→r d

(1b)

dcr ˙ D,i→r × Vr = c˙ →r − c˙ r→ + M d

(1c)

The building was considered in two structures: a heavy entirely out of concrete with additional layers of mineral wool and brick on external walls, and a light primary out of gypsum, wood and mineral wool with outside surfaces of metal sheet. Both structures represent its own extreme from a thermal characteristic point of

view and these choices were made to span the results from the investigation over most relevant cases. 2.2. HVAC systems description The HVAC system was investigated in the two variants (denoted as systems A and B) that are differentiated by their local components in Table 1. As all functions related to indoor climate control were managed with ventilation in system A, only the CO2 part was air-based in system B. Hence, both cases represent its own extreme with an extensive and minimal use of the ventilation system respectively, and these choices were made to span the results from the investigation so that most relevant HVAC systems can be found within the range. 2.2.1. Indoor climate considerations System A is an all-air HVAC system with DCV and re-heaters (RH) in each zone for temperature control, and with minimum ventilation flow levels to maintain the IAQ. System B is equipped with DCV to control the room CO2 concentrations while heating and cooling were supplied by local water-based thermal room units (TRU). All subsystems were in both cases controlled by feed-back PI-controllers with ideal parameter settings (according to the AMIGO method) and state-of-the-art sensors in the loop (with respect to accuracy and response). Identical setpoints for the local components were used throughout the study and the levels were set to ensure that the quality of the indoor climate was fulfilled at all times. In the room air temperature case, 21 ◦ C was chosen as setpoint for heating according to a national recommendation [11] which states that lower temperatures should be avoided in occupied office rooms. The setpoint for cooling was set 1 K higher, since according to the standard ISO EN 7730-2005, smaller room air temperature variations do not have a negative influence on the thermal comfort. A desirable IAQ was defined as CO2 concentrations below 1000 ppm according to several national recommendations [12] as well as the ASHRAE (American Society of Heating, Refrigeration and Air-Conditioning Engineers) standard 62-2007. This criterion was fulfilled by adjusting the CO2 level setpoints in system B and the minimum ventilation flow rates in system A. 2.2.2. Central components Throughout the study, the ventilation air consisted entirely of outdoor air that had been conditioned in a central air-handling

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

409

Table 1 Summary of local (on room level) HVAC component models used in the investigation. Local components

Description

System A

System B

Thermal room unit (TRU)

Fan-coil unit for heating or cooling. Circulating room air on one side and hot or cold water on the other. Room air side: variable drive fan. Water side: valves for constant hot or cold outlet air temperature. Small hot water-connected air-heater located directly before the supply air diffuser. Water side: valve for controlling supply air temperatures above the central setpoint. For mixing ventilation of supply and room air. Adjustable opening. Adjustable opening for balanced ventilation.

N.A.

Actuator for room air temperature.

Actuator for room air temperature.

N.A.

Actuator for room air temperature. Tracks the control signal to the supply air diffuser.

Actuator for room air CO2 concentration. Tracks the control signal to the supply air diffuser.

Re-heater (RH) of supply air

Supply air diffuser Exhaust air diffuser

Table 2 Summary of central HVAC component models. Both HVAC systems in the investigation used the same kind but with different dimensions. Central components

Description

Control arrangement

Heat recovery unit

Rotating variable speed drive non-hygroscopic wheel. Maximum temperature efficiency of 0.8. Supply air on one side and hot water on the other. Control valve for variable water flow rate. Supply air on one side and cold water on the other. Control valve for variable water flow rate. Maximum total efficiency of 0.5. Variable speed drive. Maximum total efficiency of 0.5. Variable speed drive.

First actuator in sequence for SAT control.

Air-heater Air-cooler Supply air fan Exhaust air fan

Second actuator in sequence for SAT control. Third actuator in sequence for SAT control. Flow controlled, tracks the control signals to the supply air diffusers. Flow controlled, tracks the control signals to the exhaust air diffusers.

unit (AHU) and distributed to the zones via the duct system. The central components for air-handling are presented in Table 2, and both of the considered HVAC systems used the same kind, but with different dimensions to fulfil a national recommendation of a maximum SFP (specific fan power [kW/(m3 /s)]) of 2. The first three components in Table 2 were used to control the SAT in a sequence to avoid unnecessary energy usages: the airheater could only be actuated if the heat recovery unit was in full operation, and the air-cooler could only be actuated if both the heat recovery unit and the air-heater were deactivated. The two final components in Table 2 are the central supply and exhaust air fans, and both were modelled to track the control signals of the local air diffusers in Table 1. This means that the opening of the supply air diffusers determined the ventilation air flow rates to the rooms and that balanced ventilation was achieved since the exhaust air diffusers followed the supply air diffusers.

2.2.4. Local components system B System B was introduced to consider an HVAC system in which the ventilation system only manages a minor part of the indoor climate control and therefore is used to a small extent. The room air temperatures were primary controlled by the local TRUs (Table 1), modelled as fan-coil units with the ability to either heat or cool the rooms. The IAQ was in this case maintained by directly controlling the CO2 levels by varying the openings of the supply air diffusers. The minimum flow rate was set to zero and the maximum level as dependent on the size of the room and the design number of people, both according to national guidelines [12].

2.2.3. Local components system A System A was introduced to consider an HVAC system in which the ventilation part manages all functions associated to indoor climate control and therefore is used to a large extent. The actuators for room air temperature (supply air diffusers and RHs) were controlled according to the dual maximum logic which is both efficient and relatively common in existing buildings [13]. The dual maximum logic consists of the four modes presented with different patterns in Fig. 2 as functions of the thermal power supplied by the ventilation system to the room air. For room air temperatures above the cooling setpoint (cooling supply mode), the supply air flow rate is increased while the RHs are off. Between the setpoints for cooling and heating (the dead-band mode) the RHs are still off and the supply air flow rate attains a minimum level which was set to ensure that the IAQ in the zone was maintained for the design number of people. If the room air temperature falls below the heating setpoint (and enters the first part of the heating supply mode), the RHs are actuated while the supply air flow rate is not increased until the SAT has reached its maximum (and the second part of the heating supply mode is activated).

• People • Lighting

2.3. Disturbances

Supply air flow rate

Supply air temperature Re-heat mode

Supply air flow rate

Cooling mode Heang mode Heang setpoint

Dead-band Cooling setpoint

Minimum supply air flow rate

Supply air temperature

Each simulation stretched over a working week of 5 days and nights, and was conducted by subjecting the building to time-varying sequences of the following internal and external disturbances:

Room air temperature

Fig. 2. Overview of the dual maximum control logic for the all-air HVAC system A with local re-heaters.

Equipment (such as computers and other office appliances) Opening of internal doors Solar heat gain Outdoor air temperature (OAT)

Since the corridor and the meeting room lack external walls, only the office rooms were directly influenced by the external climate and the investigation was conducted with a summer and a winter variant. Both seasons were modelled using climate data of a reference year from the Swedish coastal city of Helsingborg which has a mild tempered climate and an annual average OAT of 8.2 ◦ C. During day time, the office rooms were subjected to solar radiation from a northern (4 rooms) and a southern direction (5 rooms), and the part not blocked by the external shadings was transferred through the windows as heat. The number of people was modelled separately for each room in accordance with statistical data regarding annual occupancy in 58 rooms of a Swedish office building [14]. The outcome was a sequence in which the building could be occupied between 7:00 and 19:00 h, with normal occupancy factors (total number of occupants/maximum number of occupants) between 20 and 50%, and a maximum of 70% that occurred during one separate occasion. Further, both equipment and lighting were assumed to correlate to occupancy (i.e. light, computers and other appliances were switched on during occupied periods and switched off during vacancy). Each person was modelled to emit 18 lCO2 /h and 70 W of sensible body heat while the heat gain from lighting corresponds to 10 W/m2 according to the standard SS-EN 12464-1. The doors between the office rooms and the corridor were opened according to a stochastic variable with probability of 20% to occur during day time while the door to the meeting room was assumed to be closed at all times. 3. SAT control strategies In Table 3, the investigated SAT control strategies are summarized. Throughout, an open-loop structure was used to generate central SAT setpoints as a function of some input(s), while the room air temperature and CO2 control were managed by local HVAC components in each zone. The strategies were denoted according to their most characteristic features (which normally is the type of input signal) and can be divided in three groups depending on their purpose in the investigation. First, the OATS was considered as a benchmark and represented a conventional approach with outdoor air temperature as input. Second, the TOS, OLSS and RATS were introduced as low-complexity alternative strategies and are based on some limited information about the activities in the building. Third, the optimal strategy is extensive, has information about the entire set of disturbances and was primarily introduced to serve as an upper boundary of possible energy savings. But, the solution was moreover analysed for patterns that could be used to formulate a more general and simple strategy. All strategies were design to operate within a SAT span between 15 and 21 ◦ C. The lower boundary was set according to comfort recommendation to avoid draught and the upper boundary was the highest observed temperature for which all strategies could perform during all simulated conditions without an accelerated energy usage. The corresponding input signal values to attain these boundaries are also presented in Table 3 whenever possible. Since the focus of this study was to evaluate the performances of simple alternative strategies, it was assumed that all signals used for control were ideal. Further, no other control related energy efficiency measures were not applied during the simulations (such as relaxed comfort constrains during night time using re-circulation of ventilation air and/or reset of room air temperature setpoints).

Maximum SAT (21 °C)

(SAT) [°C]

• • • •

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

Supply air temperature

410

Minimum SAT (15 °C)

Input signal

Fig. 3. General linear and static SISO supply air temperature control strategy.

3.1. Linear SISO strategies All strategies except for the optimal are linear, static and of SISO type as presented graphically in Fig. 3. The linear representation was chosen since this form is in accordance to how the considered input signals affect the demand for heating and cooling in the building. In practice, it is possible to counteract the OATS strategy in a non-linear version but this is probably a result of a tuning process to compensate for unmeasured disturbances, such as solar radiation, which often coincide with high outdoor air temperatures. One of the main reasons for evaluating the alternative SAT control strategies was to find simpler variants to more complex control strategies such as the optimal. All of the linear SISO strategies in this investigation are regarded as non-complex due to their simple form and the measurability of their input signals. The input to the OLSS is the sum of the heating power emitted by all local units (the RHs in system A and the TRUs in system B) and a decline is interpreted as an increased cooling demand which is met by a decreased SAT. In practice, the OLSS could for example be implemented by considering Eq. (2) at each local unit. In HVAC systems like A and B, information about the water flow through each RH respective TRU would then be required together with the corresponding supply and return water temperatures. Compared to what is custom in most buildings, some additional sensors must hence be installed, but nothing outside the frames of standard products on the market. The input to the TOS is the total number of occupied rooms in the building. An increased occupancy is interpreted as an increased cooling demand which is met by a decreased SAT. This strategy could for example be realized by installing motion sensors in each space, and once again, the associated technology is then standard in building automation applications (commonly used for lighting control). In the RATS strategy, the mean room air temperature in the office and meeting rooms is used as input. This strategy only requires temperature sensors in each zone which presumably are available at most sites. Q˙ heat = V˙ water × water × cp,water × (ts − treturn ) [W]

(2)

3.2. The optimal strategy The energy used by an HVAC system to meet an instantaneous demand in a building can be optimized from two perspectives. First, an HVAC system consists of central and local components that can be operated individually or partly individually of each other, but, the operation of the central system also determines the operation of the local. If the supply from the central system is mismatched with

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

411

Table 3 Considered SAT control strategies. Name

Input signals

Input signal values to attain SAT boundaries

Outdoor air temperature strategy (OATS) Total occupancy strategy (TOS) Operation of local system strategy (OLSS) Room air temperature strategy (RATS) Optimal strategy

Outdoor air temperature.

The lowest (−11) and the highest (25) simulated outdoor air temperatures [◦ C].

Number of occupied rooms in the building.

0 and the maximum number of rooms that can be occupied (10) [–].

Sum of heating power emitted by the local units (re-heaters of TRUs).

0 and dimensioning heating power of local units times the number of rooms [W].

The mean room air temperature of the zones (not including the corridor). All internal disturbances and the outdoor air temperature.

The local setpoint for heating (21) and the local setpoint for cooling (22) [◦ C]. Multiple input and non-linear. Solution is presented further on.

the demand, the energy usage of the local components is increased. In this work, a mismatched SAT is met by an increased supply air flow rate and/or an increased power of the RHs in system A, and by an increased operation of the TRUs in system B. It is also important to emphasize that since a mismatch in one zone might be a match in another, the entire building and the variations within must be taken into account when an optimization problem is formulated. Second, dependent on the conditions, different HVAC components have different energy intensities which means that a certain cooling or heating demand can be met by using different amounts of energy. The energy intensity of the AHU is in this work dependent on the OAT and the ability to recover heat while the power of RHs and TRUs are disconnected to the external conditions. Further, the energy for central and local fans (in the AHU and TRU respectively) increases exponentially with the flow and grows faster than the energy used for air-conditioning or re-heating. The optimal strategy was derived by solving the previously described optimization problem for a static system. The heuristic solver GA was used to find the SAT that minimizes the objective function in Eq. (3a) which is a sum of all energy terms allocated to the HVAC system. Off-line solutions were returned for the considered variants of HVAC system, building and disturbances (internal and external), and in the next step, these were implemented in the simulations as the optimal strategy. This means that the optimal strategy requires information about internal and external disturbances (which naturally leads to a high level of complexity) and should therefore not be regarded as an implementable alternative in its initial form.

3.2.1. Procedure In this section, the procedure that was followed to derive the optimal strategy is presented. The building and the HVAC systems were first described as static energy balances (remark: the solution is hence optimal from a static point of view) which was done by omitting the corridor since this space is neither conditioned nor ventilated directly. The individual terms in the objective function in Eq. (3a) were then specified as functions of SAT, according to the procedure given in Eqs. (3b)–(3f). The third step was to determine the required ventilation flow rate to fulfil the indoor climate criterions in occupied rooms with heat surplus: in system A, to maintain the cooling setpoint (Eq. (3b)), and in system B, to remain below a CO2 concentration of 1000 ppm. Eq. (3c) was in turn used to specify the corresponding energy usage for air-conditioning (not including recovered heat) for a specific OAT. In the fourth step, the thermal energy usage of the local components (RHs and TRUs) was specified by Eq. (3d): in system A, as the increase of SAT needed to cover any heat deficits, and in system B, as the additional heating or cooling not covered by the ventilation system. For cases when the heating demand was not covered by the RHs in system A, the supply air flow rate was also increased to the necessary level. Eqs. (3e) and (3f) were then used to specify the associated electricity usage of central

fans and the TRUs. In the last step, the optimization problem in Eq. (3g) was formulated, and all constraints that were considered are also summarized under this index. Constraint 1–3 were embedded in the objective function (as described in the previous text) while 4–5 were managed by the solver. Most constraints (1–5), are associated to comfort considerations; the first and second state the criterions set to fulfil a satisfying indoor climate quality (see Section 2.2.1) and the third and fourth state the limits of the supply air temperature from the central and local (TRU and RH) components. The last constraint states the limitations of the air flow rate in each room (see Section 2.2).



J=

˙ AHU (ts) + |Q˙ AHU (ts)| + W

10 

|Q˙ local,i (ts)| +

i=1

10 



˙ local,i (ts) W

(3a)

i=1

Q˙ vent,i = V˙ s,i × air × cp air × (ts − tr,i )

(3b)

Q˙ AHU = V˙ s tot × air × cp air × (ts − to )

(3c)

Q˙ vent,i + Q˙ local,i = Q˙ demand,i = Q˙ external,i − Q˙ internal,i

(3d)

 ˙ AHU = W ˙ AHU max × W

 ˙ local = W ˙ local max × W

V˙ s tot ˙Vs tot max

3

Q˙ local tot Q˙ local tot max

(3e)

3 (3f)

ts central = min(J) subjected to 1. 21 ◦ C ≤ tr,i ≤ 22 ◦ C 2. cr,i ≤ 1000 ppm

(3g)

3. 15 ◦ C ≤ ts local ≤ 30 ◦ C 4. 15 ◦ C ≤ ts central ≤ 21 ◦ C 5.

V˙ s min,i ≤ V˙ s,i ≤ V˙ s max,i

3.2.2. Optimal solutions The previously described procedure was repeated for the building and HVAC system variants, and GA was used to find solutions for all considered combination of internal and external disturbances. The results are presented in Figs. 4 and 5 for HVAC systems A and B respectively, while it was found that the solutions were more or less independent of the type of building. To facilitate visualization, the optimal SAT is presented as functions of the OAT and the number of people, which is possible in this work since the influence of solar radiation could be neglected due to external shadings and since the number of people are correlated to the rest of the internal disturbances. Further, the solution for system B remained constant below an OAT of 15 ◦ C which is the reason to why the x-axis in Fig. 5 is limited.

412

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

Fig. 4. Optimal supply air temperature control strategy for HVAC system A.

In Fig. 4, the optimal strategy for HVAC system A is presented. It is best explained by dividing the solution into three regions with respect to OAT, as well as by differentiating between periods with large and small internal disturbances. The first region is associated to OATs below −8 ◦ C, and by attaining the lowest SAT boundary, energy for air-conditioning was minimized while local components were favoured in rooms with heating demands. In the second region between −8 and 19 ◦ C, the allowed SAT span between 15 and 21 ◦ C could be maintained solely by recovered heat. As the cooling demand was increased during periods with large internal disturbances, the supply air flow rate (and in turn the associated energy for fans and air-conditioning) was reduced by lowering the SAT. As the heating demand was increased during periods with small internal disturbances, the energy for RH was reduced by increasing the SAT up to the maximum temperature achievable by heat recovery. As the OAT was increased into the third region above 19 ◦ C, a more and more homogenous demand during periods with large and medium internal disturbances was met by a decreased SAT (as the transmission losses were decreased with higher OATs, even vacant

rooms got a demand for cooling due to the heat transferred from occupied rooms). During periods with small internal disturbances, on the other hand, a high SAT was favoured in order to reduce the energy associated for cooling outdoor air. The optimal strategy for HVAC system B is presented in Fig. 5. The SAT is in this case independent of the internal disturbances and consistently follows the OAT, which means the strategy can be implemented only with an external temperature sensor as input. The solution indicates that the ventilation system was disconnected from temperature control considerations by the solver: since the supply air flow rate was locked by the IAQ demand, the ventilation system was unable to provide the necessary span of heating and cooling demands within the building simultaneously, and this function was instead allocated to the TRUs. 4. Results In this work, simulations were used to investigate three lowcomplexity strategies for central SAT control in an office building,

Fig. 5. Optimal supply air temperature control strategy for HVAC system B.

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

by comparing to an optimal algorithm as an ideal case and a conventional OAT-based strategy as benchmark. Two different HVAC systems, two types of buildings and two weather seasons were considered. The results in the following text were calculated using Eq. (4) and consists of the HVAC system energy usage during two working weeks of Swedish summer and winter climate (Etotal ) respectively, weighted according to Energy Efficiency Directive 2006/32/EG. Etotal = (|Qthermal | + (Welectricity × 2.5))summer + (|Qthermal | + (Welectricity × 2.5))winter

[kW h]

(4)

In Figs. 6 and 7, the energy usage of HVAC system A is presented for the heavy and light buildings respectively. The ranking provided by sorting the alternative strategies according to their potential of saving energy compared to the benchmark (OATS) was throughout consistent. That is, the optimal strategy had the highest performance (with savings of 31 respectively 39% in the heavy and light building cases), subsequently followed by TOS (27 respectively 31%), RATS (19 respectively 29%) and OLSS (12 respectively 24%). Hence, from these figures it is clear that the energy savings achieved by replacing the benchmark with the investigated strategies were in general larger in the light building case, both in relative and absolute terms. In Figs. 8 and 9 the energy usage associated to HVAC system B is presented for the heavy and light buildings respectively. Compared to the system A cases, there are three significant differences. First, both the absolute and relative savings associated to the alternative strategies were in general lower. This is because the supply air flow rate was independent of the SAT control at the same time as a smaller part of the energy usage derived from the ventilation system. Second, also the absolute energy usages were in general lower since the ventilation system was shut down during the night (due to the absence of CO2 sources.) Third, the ranking of energy savings achieved by the alternative strategies was (although consistent) of a different order. That is, the optimal strategy still had the highest performance (with savings of 12 respectively18% in the heavy and light building cases) but was subsequently followed by OLSS (10 respectively 15%), RATS (9 respectively 13%) and TOS (6 respectively 7%). Hence, the highest energy savings potential was also in this case associated to the light building cases. 5. Conclusions and discussion The aim of this work was to find low-complexity SAT control strategies for a decreased HVAC system energy usage and a maintained indoor climate quality. Constant and identical setpoints for local components were used throughout the simulations to ensure a desirable indoor climate while the investigated strategies were generating setpoints for central SAT. The results showed that all alternative strategies resulted in lower energy usages compared to a conventional OAT-based, and in some cases substantial reductions were indicated. It was found that up to 30 and 15% of the total energy usage could be saved by applying a low-complexity strategy to HVAC system A and B respectively; compared to 39 respectively 18% when under the control of the optimal algorithm. While major energy savings with optimal strategies already have been proclaimed in previous works (e.g. about 30–40% in [4–7]), the main conclusion from this paper is that similar results can be reached with a considerable lower level of complexity. Moreover, it can also be concluded that the possible savings were highly dependent on the considered HVAC system and somewhat dependent on the considered building structures. In the following text, these aspects are first discussed which is followed by an evaluation of the alternative control strategies.

413

5.1. Influence of HVAC system and building structure In general, the energy savings potential was considerable larger in HVAC system A because the entire energy usage was allocated to ventilation, and that a major share went for air-conditioning. Further, the strategies had the possibility of decreasing all energy accounts by indirectly reducing the supply air flow rate by assigning more appropriate SATs. This means that both additional energies for air-conditioning were saved at the same time as the central fan operation was reduced. In the HVAC system B cases, only the airconditioning energy associated to SAT control could be affected by the strategies. Further, most of the thermal control was managed by the TRUs while a considerable smaller part of the energy usage derived from the ventilation system due to relatively low air flows and off-line mode during unoccupied periods. Regarding the influence of building structure, the overall savings were in general larger in the light cases, and the explanation lies in the properties of the respective building materials. Recall that the heavy structure primarily was made out of concrete and the light structure of gypsum, mineral wool and metal sheet. The outcome was a smaller thermal resistance and a larger capacity in the heavy case, which is equivalent to smaller impacts of disturbance on individual rooms. A higher capacity means that disturbances were dampened to a larger extent and a lower resistance means that heat was transferred more easily between the rooms (hence, temperature differences between occupied and empty rooms was reduced). To sum up, these two features resulted in that the light structure was more difficult to control, and in turn, the room for improvements became larger because a mismatched SAT was more penalized by an increased operation of the local HVAC components. In this context, it is also worth mentioning that the different thermal characteristics of the buildings also led to considerable higher cooling demands in light occupied rooms (visualized by the absolute scale in Fig. 9). 5.2. Evaluation of alternative strategies 5.2.1. Optimal strategy Independent of the considered HVAC systems and building variants, the optimal strategy led to the largest energy savings, and in particular when used in HVAC system A (Fig. 4). Since the optimal strategy in this case was specific and required both internal and external disturbances as inputs, it represents an upper boundary of possible energy savings and not a realistic strategy for implementation. For HVAC system B, on the other hand, the optimal strategy was general and the only necessary input signal was the OAT (Fig. 5). That is, an external temperature sensor is the only element required for implementation, which means that all other strategies are more complicated at the same time as their performances furthermore are lower in comparison. 5.2.2. SISO strategies The ranking of energy savings associated to the alternative SISO strategies was on a declining scale: TOS, RATS and OLSS for system A and OLSS, RATS and TOS for system B. In the following text, these ranking orders will be discussed and explained. All of the alternative SISO strategies utilized information about internal disturbances (either directly or indirectly) and were intended to provide a reflection of the demand within the building. The differentiation is regarding how and how fast they reacted when a shift in internal disturbances occurred. The TOS considered whether there was a cooling or a heating demand in single zones and acted directly, before the impacts of internal disturbances were observable; RATS considered whether there was a cooling or a heating demand on average and acted when the impact was seen on the room air temperatures; the OLSS utilized the operation of local

414

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

Fig. 6. The energy used by HVAC system A during two working weeks of Swedish summer and winter climate in a heavy office building. Each bar is associated to a specific SAT control strategy.

Fig. 7. The energy used by HVAC system A during two working weeks of Swedish summer and winter climate in a light office building. Each bar is associated to a specific SAT control strategy.

systems for heating and did not act until they responded. These diverse input signals led among other things to different levels of independence: while the TOS acted independently of the local systems, the RATS acted in parallel and the OLSS acted dependently. This feature explains the ability of the strategies to save energy in the two HVAC system cases. In HVAC system B, the local systems

should first and foremost manage the room temperature control while the ventilation system remains passive (according to the optimal strategy) which means that the TOS was unsuitable and the OLSS was preferable. In the system A case, the situation was the opposite and the strategy which reduced the use of local system (RHs) was most preferable (the TOS).

Fig. 8. The energy used by HVAC system B during two working weeks of Swedish summer and winter climate in a heavy office building. Each bar is associated to a specific SAT control strategy.

M. Gruber et al. / Energy and Buildings 82 (2014) 406–415

415

Fig. 9. The energy used by HVAC system B during two working weeks of Swedish summer and winter climate in a light office building. Each bar is associated to a specific SAT control strategy.

Summary

References

In this work, supply air temperature (SAT) control strategies for a reduced energy usage and a maintained indoor climate in office buildings were investigated through simulations. In total, five strategies were considered, whereof three were proposed as low-complexity alternatives, one was optimal and represented the upper boundary of possible energy savings, while the last was a conventional outdoor-air-temperature-based approach used as benchmark. Two different heating, ventilation and air-conditioning (HVAC) systems and two types of buildings were considered and the results were presented as the energy used during two working weeks of Swedish summer and winter climate respectively. It was shown that while satisfying thermal climate and indoor air quality, up to 30% of the energy usage could be saved by a low-complexity strategy while the optimal resulted in up to 39%. Moreover, it was found that the optimal strategy in one of the HVAC system cases was general and only dependent on the outdoor air temperature. This means that the solution can be used to generate a simple version while maintaining the high energy savings potential. Overall, it can hence be concluded that there is a large potential for revising SAT control, and that simple strategies are sufficient to drastically reduce the energy usage. But, it can also be concluded that the benefits were highly dependent on the considered HVAC system and somewhat dependent on the considered building structures.

[1] D. Kolokotsa, D. Rovas, E. Kosmatopoulos, K. Kalaitzakis, A roadmap towards intelligent net zero- and positive-energy buildings, Solar Energy 85 (12) (2010) 3067–3084. [2] L.K. Norford, A. Rabi, R.H. Socolow, Control of supply air temperature and outdoor airflow and its effect on energy use in a variable air volume system, ASHRAE Transactions 92 (Part 2B) (1986) 30–35. [3] N. Nassif, S. Kajl, R. Sabourin, Optimization of HVAC control system strategy using two-objective genetic algorithm, HVAC&R Research 11 (3) (2005) 459–486. [4] M. Mossolly, K. Ghali, N. Ghaddar, Optimal control strategy for a multizone air conditioning system using a genetic algorithm, Energy 34 (1) (2009) 58–66. [5] R. Parameshwaran, R. Karunakaran, C.V.R. Kumar, S. Iniyan, Energy conservative building air conditioning system controlled and optimized using fuzzy-genetic algorithm, Energy and Buildings 42 (5) (2010) 745–762. [6] S. Wang, X. Jin, Model-based optimal control of VAV air-conditioning system using genetic algorithm, Building and Environment 35 (6) (2000) 471–487. [7] F. Engdahl, D. Johansson, Optimal supply air temperature with respect to energy use in a variable air volume system, Energy and Buildings 36 (3) (2004) 205–218. [8] M. Gruber, A. Trüschel, J.-O. Dalenbäck, Model-based controllers for indoor climate control in office buildings – complexity and performance evaluation, Energy and Buildings 68 (Part A) (2013) 213–222. [9] M. Gruber, Demand-based control of indoor climate in office buildings (Licentiate thesis), Building Services Engineering, Chalmers University of Technology, 2012. [10] B.R. Sorensen, Applications and energy consumption of demand controlled ventilation systems – modelling, simulation and implementation of modular built dynamical VAV systems and control strategies (Ph.D. thesis), Department of Building Science, NTNU, 2002. [11] Belok, Energikrav för lokalbyggnader (Energy requirements for commercial buildings), Göteborg, Sweden), 2011. [12] L. Ekberg, R1, VVS-tekniska Föreningen, Kristianstad, Sweden, 2006. [13] S.T. Taylor, J. Stein, G. Paliaga, H. Cheng, Dual maximum VAV box control logic, ASHRAE Journal (December) (2012). [14] M.-L. Maripuu, Demand controlled ventilation (DCV) in commercial buildings (Ph.D. thesis), Building Services Engineering, Chalmers University of Technology, 2009.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.enbuild. 2014.06.056.